KEY TAKEAWAYS
CONTENTS
Poisson distribution betting models take one input per team, the number of goals that team is expected to score, and convert it into the probability of every possible scoreline. Those scoreline probabilities can then be added up into the prices you see on a soccer betting menu: the 1X2 (home, draw, away) market, over/under totals, both teams to score and correct score. Turning a probability into a price is the same step covered in our guide on how to convert probability to fair betting odds; what Poisson adds is a structured way to get the probability in the first place.
This matters because soccer odds are, underneath, statements about goals. A sportsbook’s home-win price, its 2.5-goal total and its 1-1 correct score price all describe the same match, so they should be consistent with one another. This article explains the formula, walks through a full worked example with hypothetical expected goals of 1.6 and 1.1, converts the results into fair decimal and American odds, and then covers the limits of the model, because a Poisson output is an estimate built on assumptions, not a prediction of what will happen.
What the Poisson Distribution in Soccer Measures
The Poisson distribution is a probability formula for counting events: how many times something happens within a fixed window, when you know the average rate. Goals in a 90-minute soccer match fit that shape well: they are countable, relatively rare, and spread through the game. The distribution needs only one number, written as the Greek letter lambda (λ), which is the expected count. A useful property is that λ is both the average and the variance of the distribution.
In betting terms, λ is a team’s expected goals for this specific match. It is not the same as a season-average goals figure, and it is not the same as a team’s xG (expected goals from shot quality) in a past game, although analysts often use both to build it. A typical approach blends a team’s attacking strength, the opponent’s defensive strength, the league’s average scoring rate and home advantage into one λ for each side.
Once each team has a λ, the model produces a probability for that team scoring exactly 0, 1, 2, 3 and so on. Those per-team probabilities are the building blocks; every market price comes from combining them. This is why Poisson sits in the Reading Odds category rather than strategy: it is a translation layer between a goals estimate and the numbers on the odds board, and it relies on the same idea of implied probability that links any price to a percentage.
How Expected Goals Become Scoreline Probabilities
The process has three steps. Each one is simple arithmetic, which is part of why the model is so widely used as a starting point.
Step 1: Apply the Poisson formula to each team
The probability of a team scoring exactly k goals is P(k) = λk × e−λ ÷ k!, where e is the mathematical constant (about 2.71828) and k! means k factorial (for example, 3! = 3 × 2 × 1 = 6, and 0! = 1). For a home side with λ = 1.6, the chance of scoring exactly two goals is 1.6² × e−1.6 ÷ 2 = 2.56 × 0.2019 ÷ 2, or about 25.84%. In plain English, if this match were played many times under identical conditions, the model expects the home side to score exactly twice in roughly one game in four.
Step 2: Combine the two teams into a scoreline grid
The basic model treats the two teams’ goal counts as independent. That means the probability of an exact score is the home probability multiplied by the away probability. The chance of a 2-1 home win is P(home scores 2) × P(away scores 1). Repeating this for every combination, usually up to six or more goals each, produces a grid of scorelines whose probabilities sum to essentially 100%.
Step 3: Add up the grid into betting markets
Every market is a group of cells in that grid. The home-win probability is the sum of every cell where the home score is higher; the draw is the diagonal (0-0, 1-1, 2-2 and so on); the away win is everything else. Under 2.5 goals adds every scoreline with two or fewer total goals. Both teams to score is one minus the chance that either side is held to zero. A correct score price simply reads one cell. Because all of these come from the same grid, they are internally consistent: the 1X2 probabilities add to 100%, and so do over plus under.
The final step is the conversion. Fair decimal odds equal 1 ÷ probability, and fair American odds follow the standard conversion: for a probability above 50%, −100 × p ÷ (1 − p); below 50%, +100 × (1 − p) ÷ p. “Fair” here means the price with no sportsbook margin included.
Worked Example: 1.6 vs 1.1 Expected Goals
Suppose a hypothetical match where your estimate is 1.6 expected goals for the home side and 1.1 for the away side, a combined 2.7 goals. These are illustrative inputs, not a real fixture. The first table shows each team’s goal probabilities from the formula above (e−1.6 ≈ 0.2019 and e−1.1 ≈ 0.3329).
| Goals scored | Home (λ = 1.6) | Away (λ = 1.1) |
|---|---|---|
| 0 | 20.19% | 33.29% |
| 1 | 32.30% | 36.62% |
| 2 | 25.84% | 20.14% |
| 3 | 13.78% | 7.38% |
| 4 | 5.51% | 2.03% |
| 5 or more | 2.37% | 0.54% |
Multiplying across gives the scorelines. The single most likely score is 1-1 at about 11.83% (32.30% × 36.62%), followed by 1-0 at 10.75%, 2-1 at 9.46%, 2-0 at 8.60%, 0-1 at 7.39%, 0-0 at 6.72% and 2-2 at 5.20%. Note that even the most likely result happens less than one time in eight. Summing the full grid produces these market prices:
| Market | Model probability | Fair decimal odds | Fair American odds |
|---|---|---|---|
| Home win | 48.96% | 2.04 | +104 |
| Draw | 24.89% | 4.02 | +302 |
| Away win | 26.15% | 3.82 | +282 |
| Over 2.5 goals | 50.64% | 1.97 | −103 |
| Under 2.5 goals | 49.36% | 2.03 | +103 |
| Both teams to score: Yes | 53.24% | 1.88 | −114 |
| Correct score 1-1 | 11.83% | 8.45 | +745 |
Reading the table: the home side is a slight underdog against the field at 48.96%, because “not home win” (draw plus away) is 51.04%. A $100 stake at the fair home price of 2.04 would return a $204 payout, or $104 profit, if the home side wins. A $10 stake on 1-1 at its fair price of 8.45 would pay out $84.50, a profit of $74.50, and would lose on any other score.
Reading Model Fair Odds Against a Sportsbook Price
The practical use of a Poisson grid is comparison. Suppose a sportsbook lists this hypothetical match at 1.95 home, 3.60 draw and 4.20 away. Their implied probabilities are 51.28%, 27.78% and 23.81%, which add to 102.87%. The 2.87 percentage points above 100% is the bookmaker’s margin, which is why sportsbook prices are normally shorter than any fair price.
Against the model, the home price of 1.95 implies more than the model’s 48.96%, and the draw at 3.60 implies more than 24.89%, so both are shorter than the model’s fair odds. The away price of 4.20 implies 23.81%, lower than the model’s 26.15% estimate. If, and only if, the model’s numbers were correct, a $100 bet on the away side at 4.20 would have a theoretical expected value of about +$9.84 (0.2615 × $320 profit minus 0.7385 × $100 loss).
That “if” carries the whole conclusion. A gap between your model and the market is either a pricing difference or an error in your inputs, and the market has usually seen information your λ values do not include: team news, tactics, weather, motivation. Sportsbooks themselves use far richer models than a basic Poisson grid, a process explained in how sportsbooks set betting odds. The more sensible reading of a large disagreement is often “check the inputs,” not “the market is wrong.” Even a genuinely positive-expectation bet still loses roughly three times in four at these probabilities, so results over a handful of bets say very little.
Limitations and Common Mistakes With Poisson Models
Independence is an assumption, not a fact. The basic model multiplies home and away probabilities as if one team’s goals had no effect on the other’s. In real matches, a team that falls behind often attacks more, and a team that leads may sit deeper, so scoring rates shift with the score.
Low-scoring draws tend to be underrated. Many analysts who test simple Poisson grids against real results find that 0-0 and 1-1 occur more often than the model predicts. That is why more advanced academic models, such as Dixon and Coles’ 1997 Poisson-based model of English league scores, add extra structure on top of the basic formula, including allowing team strength to change over time.
The output is only as good as λ. A common mistake is to plug in raw season-average goals without adjusting for opponent, venue or recent changes. Shifting the home λ from 1.6 to 1.4 moves every price on the grid, so small input errors become pricing errors across all markets at once.
Fair odds are not a prediction. A 48.96% home-win probability does not mean the home side “should” win; it means the model expects it slightly less than half the time. Treating a model edge as certain, or raising stakes because a spreadsheet says so, ignores variance and the very real chance the model is mis-specified.
Where Poisson Pricing Shows Up in Soccer Markets
You will see Poisson-style logic most clearly in markets that depend on the exact number of goals. Correct score bets are the purest example: each price is one cell of a scoreline grid, which is why 1-1 and 1-0 are usually the shortest prices and why long-shot scores carry large odds. Totals and both-teams-to-score lines, covered in our guide to soccer totals and both teams to score betting, are sums of grid cells, so they move together when expected goals change.
The three-way result market is also a grid sum, which is why the draw price is so sensitive to the total expected goals: lower combined λ means more weight on the diagonal and a shorter draw. Because the draw is a live outcome in soccer’s 3-way moneyline, a model that misjudges the diagonal misprices all three results at once. Reading the odds board with this structure in mind helps you spot when one market looks out of line with the others for the same match.
Related Concepts to Learn Next
The Poisson model sits in the middle of a learning path. Before it, you need to be comfortable with how a price expresses a percentage, the foundation of implied probability, and with removing a margin to reach a fair price. After it, the natural next step is understanding how the bookmaker’s margin is spread across a market, which our article on overround in betting odds covers with its own worked examples. From there, comparing your own fair prices against the market becomes a question of expected value and bankroll discipline, which belong to strategy rather than odds reading, and always come with the reminder that no model removes uncertainty from a single match.
Frequently Asked Questions
